Big Question 2

Tracking prediction in the brain during language processing using neural oscillations



Supervisors: Floris de Lange and Peter Hagoort
Postdoctoral Research Associate: Ashley Lewis
Starting date: July 01, 2018

(last update 2019-07-03)

Research content

The human brain actively anticipates future input to facilitate processing. This predictive capacity of the human brain is also evident in language. Neural oscillations are believed to play an important role in this process. A comprehensive understanding of the interplay between neural oscillations and their role in predictive language processing is however still lacking. This position will focus on the neural implementation of prediction in language and the role of neural oscillations therein, using neuroimaging (MEG and fMRI) methods in healthy adult volunteers. The first MEG project being developed uses different levels of sentence constraint during reading to achieve differential predictability of words of interest in a sentence.  It also employs semantic anomalies as a method of investigating semantic prediction error.  Importantly, these factors are also crossed with the reliability of the linguistic input signal, so that participants are aware that they are either in a visually degraded block or not.  The rationale is that the presence of visually degraded target words should lead participants to shift their reading strategies to rely more on top-down information based on sentence context rather than on the bottom-up information they receive by reading words.  Oscillatory signatures related to prediction and prediction error are expected to therefore be enhanced in degraded compared to non-degraded blocks.  This will provide insight into whether or not these oscillatory signatures track prediction and prediction error (rather than some other cognitive function like attention or memory processes), and potentially enable investigation of the linguistic level (semantic or word form) at which prediction errors are indexed by each oscillatory measure.

Progress 2018

Experimental design for this experiment was completed in 2018 and the project was successfully proposed and approved at the DCCN in a project proposal meeting.  Stimulus construction was completed in 2018 but still required checking by a native Dutch speaker (to take place in 2019).  Methods for stimulus degradation were explored and implemented, as well as an approach to titration of the degradedness of the target words on a participant-specific basis.  Initial preparations for use of the MEG lab etc. were completed and data collection is expected to get underway in February/March 2019.

Groundbreaking Characteristics

This project combines expertise on prediction and predictive coding in the brain and expertise on how language processing is supported by the brain to investigate what can be learned about the intersection of these two important functions by investigating neural oscillations.
The use of degraded visual input to manipulate the use of top-down compared to bottom-up reading strategies is novel and combines the expertise from researchers on the project with prediction and predictive coding backgrounds and those with language backgrounds.  Understanding how neural oscillations index sources of top-down and bottom-up information in cortical hierarchies relate to language processing is an important goal of the BQ2 and this project directly addresses this question.